Brazil’s top electoral court unanimously upheld an order on October 16 requiring Meta to treat presidential candidates equally across its platforms. The order followed a dispute over AI-generated campaign images and carries a fine of 500,000 reais per day for non-compliance, Agência Brasil reported.

How the image dispute began

Luiz Inácio Lula da Silva’s campaign complained that Meta AI refused some prompts to create images of the incumbent, while similar requests about challenger Flávio Bolsonaro worked. The complaint formed part of broader allegations that Meta’s systems were limiting Lula campaign content. The allegations have not established deliberate political bias.

Meta denied political bias. The company told the Associated Press that a filter loophole involving candidates’ generic first names and electoral numbers briefly caused the mismatch. Meta said it had taken additional steps to address the problem.

The order reaches beyond AI images

Meta must not reduce the distribution of candidate-related posts, remove recommendations or restrict searches in ways that treat the campaigns unequally. The order applies ahead of Brazil’s October 25 presidential runoff between Lula and Bolsonaro.

The court’s order covers equal treatment across posts, recommendations and searches, not just AI image generation.

The ruling puts a practical question to election platforms: can a technical filter or recommendation system affect campaigns unequally even when a company denies political intent? The court upheld the equal-treatment requirement. Meta says the image mismatch came from a technical loophole, not political preference.